Event-related fMRI is appealing for augmenting diagnosis, planning treatment, and monitoring course in clinical populations. However, fMRI must be able to accommodate variable performance success, failure to conform to task protocol, and reduced tolerance for the physical demands of an imaging session. We propose to exploit a "time aware" data acquisition and processing system to implement an interactive approach to fMRI, tailored to individual subject performance. The system allows real-time paradigm control featuring interaction of the subject with an examiner. The paradigm chosen for proof of concept is cue-facilitated word retrieval, in which aphasics perform naming to visual confrontation, and are given phonemic cues when they fail to name an item. The software architecture permits generation of statistical contrasts that take account of actual performance, in real time, and incremental monitoring of the statistical quality of the data. The goal is to generate standardized results for longitudinal studies of speech production in aphasia. We will evaluate the stability of these measures in chronic aphasics and in normal subjects. The results will provide a basis for assessing feasibility and power of future applications of fMRI in aphasia rehabilitation. The tools developed here will be made available to the research community. [unreadable] [unreadable]